You are correct. The entire ladybug_rhino package houses all of the dependencies that the Ladybug Tools plugin has on RhinoCommon. So, any time that you see that package being used, you will need to at least have Rhino Compute available as @Mathiassn says. I think McNeel charges 5 cents per compute-minute (or something like that) if you wanted to set up your own Rhino compute server. I don’t know if they’ve made it possible to run Rhino Compute on Linux.
I’ll also say that most of the heavy-lifting to correctly format Honeybee geometry for EnergyPlus is done by the ladybug_geometry package, which is pure Python and has no dependencies on RhinoCommon. The ladybug_geometry package is the fundamental reason why our HBJSON format that we use to transfer Honeybee energy models doesn’t have to obey the unbelievably restrictive rules of geometry in IDF format (eg. no windows can have more than 4 vertices, all geometry has to be ordered with counter-clockwise vertices pointed out from the room volume starting with the upper-left vertex). So you can run the ladybug_geometry package on a Linux server and that can get you most of the way there. You just have to respect the AGPL license of ladybug_geometry.
There are just two things that the ladybug_geometry package does NOT do and you will need to use a powerful compute engine like Rhino for:
- Conversion of curved geometries to planar ones (essentially all of ladybug_geometry is for planar Faces)
- Intersection of solids to form matching surfaces
And that’s a useful example for unit testing with Rhino common. I’ll use it as a reference when I write tests for the ladybug_rhino package.